Overview and calculation of X-ray K-shell transition yields for comprehensive data libraries

Funding information Conférence des présidents d'université, Grant/Award Number: Ações Integradas Luso-Francesas, Ação Ref. TC-; Conselho de Reitores das Universidades Portuguesas; European Association of National Metrology Institutes, Grant/Award Number: 17FUN02 MetroMMC; Fundação para a Ciência e a Tecnologia, Grant/Award Number: PD/BD/105919/2014, PTDC/FIS-AQM/31969/2017, Programa Pessoa 2017-2018, Proc. N 441.00 France, SFRH/BPD/94234/2013 and UID/FIS/04559/2013; Horizon 2020 Framework Programme, Grant/Award Number: 17FUN02 MetroMMC The simulation of atomic relaxation relies on data libraries with tabulated partial fluorescence yield values of radiative transitions, commonly derived from the Evaluated Atomic Data Library (EADL). However, recent studies support that the data library EADL could be improved by adopting Scofield's Hartree-Fock calculations instead of current Scofield's Hartree-Slater calculations. This work presents a bibliography overview of relevant atomic parameter values in order to verify the partial fluorescence yields presented in EADL. The references include libraries and articles, in which the atomic parameter values were theoretically calculated, experimentally measured, or obtained with semi-empirical and empirical fitting formulas. We present a comparison of total K-shell fluorescence yields and partial K-L2, K-L3, K-M2, K-M3 fluorescence yields that are either obtained directly from its references, or are derived from atomic parameters presented in these references. Additionally, we obtain comprehensive partial fluorescence yield values from the combination of semi-empirical and empirical fitting functions from different references. The comparisons performed in this work confirm that total K-shell, partial K-L2, and partial K-L3 fluorescence yield values, obtained from Scofield's Dirac-Slater calculations have better agreement with the most recent empirical values. Partial K-M2, and partial K-M3 fluorescence yield values obtained from Scofield's Dirac-Fock calculation have better agreement with the most recent empirical values. Therefore, further studies should be performed before changing the EADL data library.


INTRODUCTION
Several references can be found in literature containing values related to atomic transitions, such as: transition rates (also commonly referred as transition probability per time unit, or transition probability), ratios of emission rates (where the values may be the ratio of single or several transitions), shell fluorescence yields, fluorescence widths, Auger yields, Auger widths, etc. These values were obtained through experimental measurements, theoretical calculations, or semi-empirical fittings to the available data. Although total shell fluorescence yield and partial fluorescence yield values can generally be derived from these data, there is no comparison of comprehensive libraries with partial fluorescence yield values in the literature, possibly due to the different nomenclatures used in the literature.
Many codes and software of elemental analysis and Monte Carlo simulation of atomic relaxation (softwares based on the AXIL package, [1,2] PENELOPE, [3][4][5] PyMCA, [6] among many others) rely on the data available in literature regarding atomic relaxation, such as transition probabilities, fluorescence yields, partial fluorescence yields and radiationless transition yields. The accuracy of these codes and softwares are strongly related to the accuracy of the data used.
GEometry ANd Tracking (Geant4) toolkit [7,8] is widely employed for the simulation of the passage of the interaction of particles with matter. It includes a variety of packages for the simulation of the electromagnetic interactions of particles with matter using Monte Carlo methods. The Geant4 Low Energy Electromagnetic package [9,10] handles physics in energy ranges that extend down to 100 eV, with a detailed description of particle interaction by taking into account the atomic structure of matter. [11] To extend the use of the Geant4 to the simulation of experiments regarding X-ray, or Auger emission, the Low Energy Electromagnetic package incorporates a cascade model that handles atomic relaxation. [11] In the simulation of the atomic relaxation cascade, secondary photons or electrons are generated through radiative and radiationless transitions. The selection of the transitions is based on the respective partial transition yields and the energy of the emitted particles is calculated taking into account the atomic electron's binding energies. The transition yields and binding energies are tabulated in Geant4's library G4EMLOW and the values are derived from the Evaluated Atomic Data Library (EADL). [12] Several studies from Pia et al. [13][14][15] present comparisons between partial fluorescence yield values of the radiative transitions from EADL tabulations with the values obtained from different references, namely Scofield's Dirac-Hartree-Slater calculations, [16,17] Scofield's Dirac-Hartree-Fock calculations, [18,19] and experimental values obtained from Salem et al. fit to experimental data. [20] The result of these evaluations highlight that the radiative transition partial fluorescence yields obtained from Scofield's Dirac-Hartree-Fock calculations are globally in better agreement with the experimental values than the ones obtained using Scofield's Dirac-Hartree-Slater calculations. Pia et al. concluded [15] that EADL library do not provide the state-of-the-art radiative transition partial fluorescence yield values. Since Geant4 uses data from EADL, its revision would improve the accuracy of its simulation whenever processes with X-ray fluorescence are essential. These studies from Pia et al. [13][14][15] present a very complete comparison with detailed description of how the partial fluorescence yield values are derived for the different transitions. However, the partial fluorescence yields in these comparisons were normalized without taking into account the radiationless (or Auger) transition probabilities of the respective shell. These radiationless probabilities must be included because the partial transition yields presented in EADL library are normalized with these probabilities. Furthermore, codes and software where the simulation of radiationless transitions is relevant, use partial fluorescence yield values normalized taking into account the radiationless transitions. For these reasons, a comparison of comprehensive values of partial fluorescence yields obtained from different references is of significant importance for the context of improving data libraries. Such comparison can also complement the studies of Pia et al. [13][14][15] We thus present an overview of the existing references that contain comprehensive values related to atomic transitions. Furthermore, we calculate and compare K-shell fluorescence yield, and partial fluorescence yield values using the data presented in those references.

THEORETICAL BACKGROUND
An atom in an excited atomic state may decay to a lower energy level through spontaneous emission. In this process, an atomic electron spontaneously transits from an energy level E 2 to a lower energy level E 1 . The probability A of this spontaneous transition has units of s −1 . This quantity is also usually designated as "transition rate" or just "transition probability per unit of time". To avoid terminology confusion, we will refer to this quantity as "transition probability". Since the energy width Γ and the mean life of the excited atomic state are related through Heisenberg uncertainty principle Γ × = ℏ, the transition probability of the state is therefore given as A = 1∕ = Γ∕ℏ. [21] The system may transit from an excited state to lower energy states through a radiative transition, or radiationless transition. In both cases a vacancy in an atomic shell is filled by an atomic electron coming from a higher energy shell or subshell. While in radiative transitions this results in the emission of an X-ray photon, in radiationless transitions, another atomic electron of an higher energy shell or subshell is emitted instead. If the electron filling the vacancy comes from a higher shell, the radiationless transition is commonly denominated as Auger transition, and the emmited electron is the Auger electron. Specific cases of radiationless transitions include the Coster-Kronig transition, in which the electron transition is intra-shell, and super Coster-Kronig transition, where, in addition to the previous case, the Auger electron is emitted from the same shell.
For an atomic state with an initial hole in the i sub-shell, the total radiative width Γ (TR) i is the energy width associated with all radiative transitions, and the total radiationless width Γ (TA) i is the energy width associated with all radiationless transitions (including Auger, Coster-Kronig, and super Coster-Kronig transitions). The total state width is given as: . Likewise, the total transition probability of the state is given as i is the total radiative transition probability, and A (TA) i is the total radiationless transition probability. Both A (TR) i and A (TA) i are given as, where A (R) i is the transition probability of the radiative transition where the hole in i sub-shell is filled with an electron from the j sub-shell, and A (A) i k is the transition probability of the radiationless transition where the hole in i sub-shell is filled with an electron from the j and an Auger electron is emitted from the k sub-shell. The fluorescence yield of an atomic state, , is the probability that it deexcites through radiative transition. For an atomic state with a hole in the i sub-shell it is given by: or, The partial fluorescence yield of a specific radiative transition (or set of transitions) can be calculated in a similar way. The partial radiative transition fluorescence yield ij of a the transition ij, in which the hole in a i sub-shell is filled with an electron from the j sub-shell, with respective transition probability A (R) i , is written as: representing the probability that the atom will deexcite through the specific transition ij instead of all other possible radiative and radiationless transitions. As expected, for any excited state, the sum of all partial radiative transition fluorescence yields equals the fluorescence yield of the shell: Frequently, in literature, [13][14][15] the radiationless transition rates A (A) ijk are not included in Equation (4), as it was previously mentioned. The values calculated in this way have varying designations among the literature.
In Pia et al. work [13][14][15] the designation "transition probability" was used. In the present work we use the designation "partial fluorescence yield (normalized without accounting radiationless transitions)" and use the symbol " (NA) i " (where NA stands for "No Auger") to represent it. Thus, the radiative transition partial fluorescence yield (normalized without accounting radiationless transitions) " (NA) t ", for the radiative transition ij, is given as: representing the probability that the atom will deexcite through a specific radiative transition ij instead of all other possible radiative transitions. The sum of all partial fluorescence yields (normalized without accounting radiationless transitions) is unity: For any radiative transition ij, the respective partial fluorescence yield ij , and the respective partial fluorescence yield (normalized without accounting radiationless transitions) (NA) i , follow the relation: where i is the fluorescence yield of the respective sub-shell. Usually in literature, the fluorescence yield of a shell is defined in a way related to its experimental measurement. The fluorescence yield for the K shell is defined by the ratio: w K = I K ∕n K , where I K is the average number of characteristic K-shell X-rays emitted from the sample as result of radiative transition to fill the n K number of primary K-shell vacancies created in the sample. The definition of fluorescence yield for higher atomic shells is more complicated since vacancies can now be distributed in different subshells. In these cases an average or mean fluorescence yieldw is defined, the relative numbers of primary vacancies in each subshell must be taken into account. Bambynek et al. [22] proposed an unambiguous definition of the average fluorescence yieldw i for an i shell expressed as a linear combination of the relative numbers of primary vacancies in the subshells. The quantities presented in this section have various designations in literature. In fact, the various references that are presented in Section 3 often use different designations and symbols than the ones used in the present work. To help readability, in Tables 1, 2 and 3,  we compare the different designations and symbols used  in the present work, with the ones used in the references  presented in Section 3. To simplify, the variable subscripts  corresponding to the shells and sub-shells are omitted. Also to simplify, the transition probability A, presented in the first column of Table 1, can refer to either the transition probability A (R) i of a radiative transition ij, or the transition probability A (A) i k of a radaitionless transition ijk (both quantities are presented in Equation (1)). We choose to make this simplification because most of the works that are listed in Table 1 do not use different symbols or nomenclature to differ a single radiative transition probability from a single radiationless transition probability.

Atomic structure calculations
Before we present the atomic parameters from different works, it is worthwhile to present the basic description of the calculation of such parameters. All state of the art atomic structure calculations aim to solve the multi-electronic atomic Hamiltonian, where the the electron-electron interactions must be taken into account. The relativistic Hamiltonian for N-electron atomic systems can be written by: where G(i, j) is the operator for the relativist interaction between the electron i and the electron j, and H D (r i ) is the one-electron Dirac's Hamiltonian. H D (r i ) is given as: where r is the electron position, m the electron rest mass, p the momentum operator, e the Dirac matrices and V(r) the nuclear potencial. The operator G partially accounts for the relativist interaction between electrons, and it is written as: where the first term accounts for the Coulomb interaction, the second term (the Breit term) accounts for magnetic interaction and retardation, and r ij = |r i − r j |. In non-relativist methods, operator G only includes the Coulomb interaction term. Hartree-Slater and Hartree-Fock (or Hartree-Slater-Fock) methods were frequently used to achieve the solution of the potential, the energy, and the wave functions of multi-electronic systems using the variational principal to the expected value of the Hamiltonian. Nowadays, these methods have widely been replaced by their relativist versions, the Dirac-Hartree-Slater (often named Dirac-Slater) method, and the Dirac-Hartree-Fock (often named Dirac-Fock)
With this inclusion, Hartree-Fock accounts for more of the electron-electron interaction than the Hartree-Slater methods, and have generally been shown to be more accurate. Even so, the Hartree-Fock (and Dirac-Hartree-Fock) approaches do not fully describe the Coulomb repulsion between electrons. To obtain high precision results the electronic correlation needs to fully be taken into account, by using methods such as the Multiconfiguration Dirac-Fock, in which the electronic correlation is included by writing the wave function of the atomic system as a linear combination of configuration wave functions , which are wave functions for different possible configuration states: coefficients. This method follow a similar procedure as the Hartree-Fock method. The multiconfiguration Dirac-Fock method is described in detail in other works. [31][32][33][34]

OVERVIEW OF THE VALUES FROM DIFFERENT WORKS
We present references containing K-shell fluorescence yields, transition probabilities, partial fluorescence yields, or other atomic parameters, from which K-shell fluorescence yield values K and partial fluorescence yield values K-L 2 , K-L 3 , K-M 2 and, K-M 3 can be obtained, whether they are libraries, articles, product of theoretical calculations, experimental measurements or the result of combined theoretical and experimental values. Although a huge number of theoretical calculations and experimental results can be found in literature, we selected and described the ones from which comprehensive values can be obtained for large range of transitions and large range of atomic number, which are the ones relevant to the comparisons made in the present work.
The nomenclature of the atomic parameters in the references presented in this section may differ from the nomenclature in the present work. Throughout this section we refer to each quantity using the designations of the present work. The designations used in the original references are presented in Tables 1, 2 and 3.

Hartree-Slater radiationless transition probability calculations and Scofield's Dirac-Slater radiative transition probability calculations
In 1971, Kostroun et al. [23] calculated nonrelativistic K-shell radiationless transition probabilities for selected elements using the Hartree-Slater method (using the non-relativist versions of Equations (9), (10) and (11). In order to derive theoretical total K-shell fluorescence yields, Kostroun et al. [23] combined their results with Scofield's earlier Dirac-Slater radiative transition probabilities, [16] which were calculated without considering the finite extent of the nuclear charge distribution. In Kostroun et al. work, calculations where only performed for selected elements, the values for the remaining elements where obtained through fits.
In the present work, we present the K-shell fluorescence yield values K extracted directly from Kostroun et al. work. These values are compared with values from other works in Section 4, and presented in Table A1 (Section VII). Throughout this work we refer to these values as "Kostroun".

Scofield's radiative transition probability calculations based on the Dirac-Slater method
In 1969, using relativistic Hartree-Slater theory, or Dirac-Slater theory (as presented in Section 2.A), Scofield calculated K-and L-shell radiative transition probabilities [16] for selected transitions, and selected elements in the range 13 < Z < 92. In these calculations, the electrons were treated relativistically and the effect of retardation was included (the second term of operator G from Equation (9)). The electrons were considered as moving independently with their mutual interactions accounted for by a central potential (thus not including the exchange potential from Equation (11)). All multipoles of the radiation field and all transitions from occupied states of the atom were included. Scofield presents the radiative transition probabilities calculated using this model for Kand L-shell transitions. [16] As mentioned in Section 3.A, the calculations in this work did not considered the finite extent of the nuclear charge distribution.
In 1974, Scofield presented K-and L-shell radiative transition probabilities calculated using an improved model, by including nuclear charge distribution of finite extent. [17] In this later work, it is presented values for elements with atomic number ranging 5 ≤ Z ≤ 104 and for all possible radiative transitions.

Scofield's radiative transition probability calculations based on the Dirac-Fock method
Using relativistic Hartree-Fock theory, or Dirac-Fock theory (as presented in Section 2.A), Scofield calculated Kand L-shell radiative transition probabilities. In Scofield's work, the exchange potential V ex was included (as in Equation (12)). In his work, the values for K-shell are presented for selected transitions, and for selected elements with atomic number ranging 10 ≤ Z ≤ 98. [18] The values for L-shell radiative transition probabilities are presented for selected elements with atomic number ranging 18 ≤ Z ≤ 94. [19] In other works, [35] it is presented L-shell radiative transition probabilities for more selected elements, which were obtained by interpolation of the probabilities calculated by Scofield.

Chen et al. radiationless transition probability calculations based on the Dirac-Fock method
Chen et al. theoretical approach consists on calculating the radiationless transition probabilities from relativistic perturbation theory, for frozen orbitals, in the Dirac-Hartree-Fock (or Dirac-Fock) approach (thus including the exchange potential in Equation (12)). Using this approach, Chen et al. present calculated radiationless transition probabilities for K-and L-shell, for selected elements with atomic number in the range 18 ≤ Z ≤ 96. [24] In a later work, Chen et al present radiationless K-shell atomic level widths [25] and present radiationless L-shell atomic level widths. [36,37] for selected elements.

EADL
The EADL library [12] contains partial fluorescence yield and radiationless transition yield values, which the Geant4 includes in its library. The radiative transition partial transition yield values are based on Scofield's theoretical approaches [16,17] (which were presented in Section 3.A and Section 3.B) and the radiationless transition yield values are based on Chen's theoretical approaches [24,25,[36][37][38] (which were presented in Section 3.D) complemented by Hubbel's corrections [12] to avoid the over-prediction of the strength of Coster-Kronig transitions resulting from Dirac-Hartree-Slater calculations. The 2014 version of EADL can be found online [39] Recent changes have been performed to EADL library, regarding the binding energy values.
In the present work, we present K-shell partial fluorescence yield values K-L 2 , K-L 3 , K-M 2 , K-M 3 , extracted directly from the EADL library. [12] We also present K-shell fluorescence yield values K obtained by summing over all radiative transition fluorescence yields, as in Equation (5). These values are compared with values from other works in Section 4, and presented in Section VII (in Tables A1  to A5). Througouht this work we refer to these values as "EADL".

Salem et al. experimental Kand L-shell X-ray transition ratios
Salem et al. compiled selected experimental data for K-and L-shell transition rates, and through least-squares fitting, generated most probable values of K-and L-shell X-ray radiative transition rate ratios. [20] From these ratios, partial fluorescence yield values (normalized without accounting radiationless transitions) (NA) i can be obtained, as was performed in Pia et al. works, [13][14][15] and in NIST's Fundamental Parameters Database. [28]

Bambynek et al 1972 semi-empirical fittings to experimental data
In 1972, Bambynek et al. [22] fitted a collection of selected K-shell fluorescence yield experimental values, using a semi-empirical fitting formula. The formula and the fitting parameters are presented in Hubbell et al. comparisons of several compilations. [27] We present Bambynek et al. [22] K-shell fluorescence yield values K , which where extracted directly from Hubbell et al. work. [27] These values are compared with values from other works in Section 4, and presented in Section VII (Table A6).
Throughout the work, we refer to the values obtained in this subsection as "Bambynek1972".

Krause's 1979 compilation
Krause generated a consistent set of values of K-and L-shell radiative and radiationless yields using the information available up to 1979 on several atomic parameters (fluorescence yields, auger yields, transitions probabilities, level withs, etc). [40] In Krause's work, all pertinent data available in literature, including experimental, theoretical and semi-empirical values, was compiled and evaluated in order to generate K-and L-shell radiative and radiationless yields.
We present the K-shell fluorescence yield K values from Krause's compilation. [40] The values are extracted directly from Hubbell et al. comparisons of several compilations. [27] These values are compared with values from other works in Section 4, and presented in Section VII (Table A6).
Throughout the work, we refer to the values obtained in this subsection as "Krause1979".

Bambynek et al 1984 semi-empirical fittings of experimental data
In 1984, Bambynek et al. presented a new evaluation [41] of K-shell fluorescence yield values by introducing about 100 new measurements to the  work. [22] As such, new fitting parameters where obtained, and new fitted values. Hubbell et al. comparisons of several compilations [27] present Bambynek's new fitting parameters, new fitted values, and comparisons against Bambynek 1972 [22] fitted values.
We present Bambynek et al. K-shell fluorescence yield values K , which where extracted directly from Hubbell et al. comparisons of several compilations. [27] These values are compared with values from other works in Section 4, and presented in Section VII (Table A6).
Throughout the work, we refer to the values obtained in this subsection as "Bambynek1984".

Hubbell et al. 1994 semi-empirical fittings of experimental data
In 1994, Hubbel et al. compiled the measured K-shell, L-shell, and higher atomic shell X-ray fluorescence yield data generated between the period 1978 -1993. [27] From the compiled data K-, L-and M-shell fluorescence yields were produced by fitting the selected data (the selection criteria and fitting procedure is explained in their work). They compared their fitted values with earlier fitted values and theoretical values. Hubbel et al. fits for K-shell fluorescence yields K are presented in NIST Fundamental Parameters Database [28] even though an erratum [42] informs about extensive anomalies to their fitted values, and as such they advise that the values obtained from Bambynek's 1984 work [41] should be used instead, regarding K-shell fluorescence yield values.

NIST Fundamental Parameters Database
Elam accomplished the compilation of a comprehensive database of atomic fundamental parameters relevant to X-ray spectroscopy, [28] which is available in the NIST's web-page. The radiative K-shell fluorescence yield K , tabulated in the database were calculated from the fits to several experimental values by Hubbel et al., [27] discussed in the previous subsection. The radiative L-shell fluorescence yield tabulated in the database are from Krause's revision to experimental and theoretical values [40] with modifications proposed by Jitschin. [43] The database also presents partial fluorescence yield values (normalized without accounting radiationless transitions) (NA) for K-and L-shell transitions that were calculated using the emission rate ratios presented by Salem et al work [20] ; the fit process to the data is explained with detail in Elam et al.'s work. [28] We present K-shell fluorescence yields values K , extracted directly from the NIST Fundamental Parameters Database. [28] These values are compared with values from other works in Section 4, and presented in Section VII (Table A6). Throughout the work, the values obtained in this subsection are referred to as "NIST".

Combination of Chen et al. Dirac Fock transition probabilities and Scofield's Dirac-Slater transition probabilities
In the present work, we calculated K-shell fluorescence yield values K and K-shell partial fluorescence yield values K-L 2 , K-L 3 , K-M 2 and, K-M 3 , using the same references as EADL, i.e. we use Scofield's more recent Dirac-Hartree-Slater calculations of radiative transitions probabilities [17] (which were presented in Section 3.B) and Chen's Dirac-Hartree-Fock calculations of radiationless transition probabilities (which were presented in Section 3.D).
Two different works from Chen et al. [24,25] (presented in Section 3.D) can be used to obtain the K-shell total radiationless transition probability A (TA) K . Chen et al. later work presents the K-shell total radiationless width Γ (TA) K , from which the K-shell total radiationless transition probabil-ity A (TA) K can be derived. The total radiationless transition probability A (TA) K calculated in Chen et al earlier work [24] differs from the calculated value in a later work. [25] Nevertheless, the K-shell fluorescence yield values calculated using the different references have deviations of less than 0.1%, and as such, both references are equivalent. We opted of using Chen et al. later work. [25] Since Chen et al. present values for selected elements we derived the values for other elements through fits to the available data.
Thus, we calculate the K-shell fluorescence yields K and radiative transition partial fluorescence yield values K-L 2 , K-L 3 , K-M 2 and, K-M 3 using Equation (3) and Equation (4), respectively, where the radiative transition probabilities A (R) K , and the total radiative transition probability A (TR) K is obtained from Scofield's later Dirac-Slater work, [17] and where the total radiationless transition probability A (TA) K is obtained from Chen et al. later Dirac-Fock work. [25] This combination with Scofield's transition probabilities to derive K-shell fluorescence yield values K had already been performed in Chen et al. later work. [25] Unlike EADL, we don't include Hubbel's corrections [44] in any calculations since the procedure of how the EADL library applies these corrections is unclear.
The values obtained in this subsection are compared with values from other works in Section 4, and presented in Section VII (Tables A1 to A5).
Throughout the work, we refer to the values obtained in this subsection as "DF,DS", since they are a combination of Dirac-Fock and Dirac-Slater calculations.

Combination of Chen et al. Dirac Fock transition probabilities and Scofield's Dirac-Fock transition probabilities
We calculate K-shell fluorescence yield values K and K-shell partial fluorescence yield values K-L 2 , K-L 3 , K-M 2 and, K-M 3 from the combination of Scofield's radiative transitions Dirac-Hartree-Fock calculations [18] (presented in Section 3.C) and Chen et al. later radiationless transitions Dirac-Hartree-Fock calculations [25]  In Scofield's Dirac-Fock work, [18] it is presented radiative transition probabilities A (R) for a limited number of transitions, and limited number of elements, as well as total radiative transition probability A (TR) . From this information, the partial fluorescence yields (normalized without accounting radiationless transitions) (NA) can be obtained using Equation (6). As far as we know, the xraylib database [26] uses this method, with the exception that the values had to be normalized to satisfy Equation (7). This method is limited to those transitions which are presented in Scofield's work. Pia et al. performed a different approach to obtain (NA) values from Scofield's Dirac-Fock work. [18] Since Scofield's work also present radiative transition probability ratios, Pia et al. wrote equations to extract the (NA) values from these ratios. Since no transition probability ratio relative to K-L 4,5 transitions is presented in Scofield's Dirac Fock work, Pia et al. made an approximation, by assuming that the K-M 4,5 /K-M 2 Dirac-Fock ratio should be equal to the experimental K-M 4,5 /K-M 2 ratio of Salem et al. work. [20] This approximation was justified by the fact that Scofield's Dirac-Slater K-M 4,5 /K-M 2 probability ratio was in good agreement with the experimental ratio, and as such, the Dirac-Fock ratio, although not presented, should also be in good agreement. In the present work, we follow the method employed by works [13][14][15] ) produces values that are overall in slightly closer agreement with the experimental values from Salem et al. work [20] than the values obtained from the xraylib method.
We calculated the K-shell fluorescence yield K using Equation (3), where the total radiationless transition probability A (TA) K is from Chen et al. Dirac-Fock work [25] (presented in Section 3.D), and the total radiative transition probability A (TR) K is from Scofield's Dirac-Fock work [18] (presented in Section 3.C).
The partial fluorescence yields K-L 2 , K-L 3 , K-M 2 , K-M 3 are calculated from (NA) These values obtained on this subsection are compared with values from other works in Section 4, and presented in Section VII (in Tables A1 to A5). Throughout this work, we refer to these values as "DF", as they are a result of Dirac-Fock calculations.

Multiconfiguration Dirac-Fock values
We present K-shell fluorescence yield values K calculated using the Multiconfiguration Dirac-Fock method, for selected elements. The values were calculated by different authors. [45][46][47][48][49] These values are compared with values from other works in Section 4, and presented in Section VII (Table A1). Throughout the present work we refer to these values as "MCDF".

Kahoul et al. 2012 empirical fittings
In 2012, Kahoul et al. [29] compiled experimental data of the K-shell fluorescence yields from the period 1960-2011. From this data, they deduced empirical K-shell fluorescence yield values K from polynomial fittings of the weighted-mean and unweighted-mean values of all the compiled data. The data was separated in three atomic number ranges, 11 ≤ Z ≤ 20, 21 ≤ Z ≤ 50 and 51 ≤ Z ≤ 99, and a polynomial fitting was performed independently for each range.
In  (Table A6). These values are referred to as "Kahoul2012", throughout the present work.

Daoudi et al. 2015 empirical fittings
In 2015, Daoudi et al. [30] compiled experimental data of the K-shell fluorescence yield values up to the date, relying on existing compilations, and adding new data published in the period 2012-2015. This compilation included 737 experimental values, with atomic number range 3 ≤ Z ≤ 99. The data was separated in three atomic number ranges, 11 ≤ Z ≤ 20, 21 ≤ Z ≤ 50 and 51 ≤ Z ≤ 99, and an empiric polynomial fitting was performed independently for each range.
We present the K-shell fluorescence yield values K obtained from Daoudi et al. [30] empirical fittings. The values are extracted directly from Daoudi et al. work, [30] and are compared with values from other works in Section 4, and presented in Section VII (Table A6). Throughout this work we refer to these values as "Daoudi2015".

Combination of semi-empirical or empirical K-shell fluorescence yield values with Salem et al. values
From the references regarding semi-empirical or empirical fittings presented so far ("Bambynek1972 [22] ", "Krause1979 [40] ", "Bambynek1984 [41] ", "NIST [28] ", "Kahoul2012 [29] " and "Daoudi2015 [30] "), semi-empirical and empirical values regarding K-shell fluorescence yield K were obtained. In the present work, we calculate partial fluorescence yields K-L 2 , K-L 3 , K-M 2 , K-M 3 for each of these references using using Equation (8), where the partial fluorescence yield values (normalized without accounting radiationless transitions) (NA) are those obtained from Salem et al. work, [20] which are tabulated in NIST Fundamental Parameters Database. As an example, to derive the K-L 2 partial fluorescence yield K-L 2 from Daoudi et al.'s K-shell fluorescence yield K , we write: We use Salem et al. values because they are still up to date considered the most complete references of K-and L-shell experimental transition probability ratios, and their relevance is reinforced by the fact that they are used in the NIST Fundamental Parameters Database, and the fact they are used in Pia et al.'s comparisons.
Using this formalism, we present partial fluorescence yield values K-L 2 , K-L 3 , K-M 2 , K-M 3 , for each of the semi-empirical and empirical values.

COMPARISON OF VALUES FROM THE DIFFERENT WORKS
We present a comparison of the K-shell fluorescence yield values K , and the K-shell partial fluorescence yield values K-L 2 , K-L 3 , K-M 2 , K-M 3 , that were obtained from the different references as explained in Section 3.

K-shell Fluorescence Yield Comparison 4.1.1 Theoretical K-shell fluorescence yield values
The K-shell fluorescence yield theoretical values are presented in Figure 1a,b (the values are divided in two atomic ranges to better visualization), and in Section VII ( Table A1). The same theoretical K-shell fluorescence yield values are normalized to their corresponding "DF" values and presented in Figure 1c.
From the comparisons in Figure 1a,b it is highlighted that the K-shell fluorescence yield values increase monotonically with increasing atomic number for all theoretical references. While "Kostroun" values are in relatively good agreement with "DF,DS" values for low atomic numbers, their disagreement increases as atomic number increases. The "EADL" values are in good agreement with "DF,DS" values, which is to be expected since they are calculated using the transition probabilities from the same references (with the excetion that the EADL library introduces a correction that we did not introduce in "DF,DS" calculations). "DF,DS" and "DF" values are obtained using the same reference for radiationless transition probabilities, but using different references from Scofield regarding the radiative transitions probabilities. Scofield's radiative transition probabilities calculated using the Dirac-Fock method are overall higher than those calculated using the Dirac-Slater method. As such, "DF" K-shell fluorescence yield values are overall higher than "DF,DS" values, as shown in Figure 1c. For low atomic numbers, "DF,DS" and "DF" values differ by almost 10% difference, but as the atomic number increases they become more and more in agreement. In fact, for 42 < Z < 74 they differ less than 1%, and from 75 < Z < 96 they differ less than 0.1% The "MCDF" values are in better agreement with "DF" values among other theoretical values.

Semi-empirical and empirical K-shell fluorescence yield values
The semi-empirical and emperical K-shell fluorescence yield values are compared in Figure 2a,b (the values are divided in two atomic ranges to better visualization). The same values are presented in Table A6 (in Section VII). Table A6 also presents, for each atomic number, the minimum and the maximum value from the compared references, and the relative difference from those values, with the intent of presenting the highest deviations of the available semi-empirical and empirical references. In Figure 2c,d, the semi-empirical and empirical values are normalized to the corresponding theoretical "DF" values, and "DF,DS" values, respectively.
Unlike the theoretical values presented in Figure 1, some of the semi-empirical and empirical values presented in Figure 2a,b exhibit non-monotonically behaviour with increasing atomic number. Such is highlighted in Figure 2b zooms, where it is visible that in the range 70 < Z < 85 NIST's values decrease as atomic number increases, and that "Daoudi" values decrease for Z = 50. The decrease in "Daoudi" values can be explained by the fact that they divided the experimental results in three ranges of atomic number and performed different fittings for each range. From Figure 2b (and Table A6) it is visible that in the range 60 < Z < 80 "NIST" values are in clear disagreement with all other semi-empirical and empirical values. The different semi-empirical and empirical K-shell fluorescence yield values present worse agreement between each other for low atomic numbers. In the range 3 < Z < 18, differences higher than 10% can be found between some of the references. As the atomic number  [25] and Scofield's radiative Dirac-Slater calculations. [17] "DF": values obtained from the combination of Chen et al. radiationless Diroc-Fock calculations [25] and Scofield's radiative Dirac-Fock calculations. [18] "EADL": values obtained from the Evaluated Atomic Data Library. [12] Scofield's radiative Dirac-Fock calculations. [18] "EADL": values obtained from the Evaluated Atomic Data Library. [12] [Colour figure can be viewed at wileyonlinelibrary.com] FIGURE 4 "Bambynek1972": Values from Bambynek et al. earlier semi-empirical fit. [22] "Krause1979": values from Krause's semi-empirical fit. [40] "Bambynek1984": values from Bambynek et al. later semi-empirical fit [41] . "NIST": semi-empirical fit values presented in NIST's Fundamental Parameters Database. [28] "Kahoul": values from Kahoul et al. empirical fit. [29] "Daoudi": values from Daoudi et al. empirical fit. [30] increases, better agreement is generally found between the different references. Such is to be expected, since experimental measurement of K-shell fluorescence yields for low atomic numbers generally result in higher uncertainties due to the higher predominance of radiationless transition in these elements. In the range 19 < Z < 50 the difference between the different values is comprised between 10% to 1%, and in the range 51 < Z < 99 the agreement is comprised between 1% to 0.4%. Figure 2c highlights that "Bam-bynek1984" values are in excellent agreement with "DF" values, their relative difference is always lower than 0.2%. "Bambynek1972" and "Krause1979" values are also in good agreement with the "DF" values, especially in the Z > 40 range, where they converge to the "DF" values as the atomic number increases. The more recent empirical and semi-empirical values, "NIST", "Kahoul2012", and "Daoudi2015", present worse agreement with "DF" values empirical fit. [30] [Colour figure can be viewed at wileyonlinelibrary.com] than "Bambynek1972", "Krause1979", "Bambynek1984" values. In the range 0 < Z < 40 the semi-empirical and empirical values are generally lower than "DF" values. The "NIST" values present the worst agreement with the "DF" values.

The comparison in
From Figure 2d one can see that "Bambynek1972", "Krause1979" and "Bambynek1984" values are in good agreement with "DF,DS" values, and their agreement increases with increasing atomic number. "Kahoul2012" and "Daoudi2015" values are also in good agreement but their values do not converge to those of "DF,DS" values as the atomic number increases. As for the "NIST" values, they present as the ones with worse agreement with "DF,DS" values.
It is worthwhile mentioning that from the comparisons presented in this subsection, the K-shell fluorescence yield values from "DF,DS", "DF", "EADL", "Bambynek1972", "Krause1979" and "Bambynek1984", converge as atomic number increases, presenting very good agreement in the range Z > 40, while the most recent values "NIST", Kahoul" and "Daoudi2015" do not exhibit this behaviour.
In Table 4 the average relative difference between each semi-empirical and empirical values and the "DF,DS" and "DF" values is presented. From this Table it is shown that the "Bambynek 1972" and "Bambynek 1984" values are in closer agreement with the "DF" values, while all the other references are in better agreement with the "DF,DS" values. Although, with the exception of "NIST" values, all values are in good agreement with the "DF" and "DF,DS" values, presenting average relative difference lower than 2%. The "Bambynek 1984" values are in excellent agreement with "DF" values, presenting less than 0.2% difference.

K-L 2 and K-L 3 partial fluorescence yield comparison 4.2.1 Theoretical K-L 2 and K-L 3 partial fluorescence yield values
The theoretical partial fluorescence yields K-L 2 and K-L 3 of "DF,DS", "DF", and "EADL" values are compared in Figure 3a,b (values are divided in two ranges of atomic number to better visualization), and presented in Tables A2 and A3 (in Section VII). In Figure 3c,d, the "DF,DS" and "EADL" partial fluorescence yields K-L 2 and K-L 3 are normalized to the "DF" values. From Figure 3a,b (and Tables A2 and A3) it is highlighted that "DF,DS" partial fluorescence yields K-L 2 and K-L 3 are lower than "DF" values in the range 18 < Z < 36, and higher than "DF" values in the range 36 < Z ≤ 97. For low atomic numbers the difference between "DF,DS" and "DF" references is significantly high for both K-L 2 , and K-L 3 , the difference is close to 7% for Z=18. For high atomic numbers better agreement is shown, in the range Z > 60 the difference is always lower than 1%.

Semi-empirical and empirical K-L 2 and K-L 3 partial fluorescence yield values
In Figure 4a,b, the semi-empirical and empirical K-L 2 and K-L 3 partial fluorescence yield values are compared (the values are divided in two ranges of atomic number to better visualization), which are obtained from the combination of semi-empirical or empirical K-shell fluorescence yield values with Salem et al. transition probability values (as explained in Section 3.Q). The same values are presented in Tables A7 and A8 (Section VII). In Figure 4c,d, K-L 2 and K-L 3 values of semi-empirical and empirical references are normalized to the respective "DF" values, and, in Figure 4e,f, are normalized to the respective "DF,DS" values.
The values presented in Figure 4a,b (also presented in Tables A7 and A8) exhibit significantly high relative differences from each other, especially at low atomic numbers. In the range 3 ≤ Z < 20 differences of 10% to 100% are found between some of the references. As the atomic number increases better agreement is observed. In the range 20 < Z < 80, the values have relative differences within 10%, and in the range 80 < Z < 96 lower than 2%.
The comparisons in Figure 4c,d, 4.e and 4.f, show that at low atomic numbers, better agreement is seen in the comparison with "DF,DS" values, with the exception of the "Bambynek1984, Salem" values, which are in better agreement with "DF" values. For higher atomic numbers the semi-empirical and empirical values are generally in better agreement with "DF" values.    [25] and Scofield's radiative Dirac-Slater calculations. [17] "DF": values obtained from the combination of Chen et al. radiationless Diroc-Fock calculations [25] and Scofield's radiative Dirac-Fock calculations. [18] "EADL": values obtained from the Evaluated Atomic Data Library. [12] Table 5 presents the average relative difference between each semi-empirical or empirical K-L 2 values and the "DF,DS" and "DF" values. In Table 6 the same is presented, but regarding K-L 3 values. From these Tables, we infer that all semi-empirical and empirical values are in closer agreement with "DF,DS" values than the "DF" values, with the exception of "Bambynek1984, Salem" (for both K-L 2 and K-L 3 ), and "Kahoul 2012" (for K-L 2 ). It is worthwhile noting that the average deviation between the semi-empirical and empirical values in relation to "DF,DS" or "DF" values is for all cases lower than 3%, and in many cases lower than 1%.

K-M 2 and K-M 3 fluorescence yield comparison 4.3.1 Theoretical K-M 2 and K-M 3 partial fluorescence yield values
A comparison of the theoretical partial fluorescence yield values K-M 2 and K-M 3 is presented in Figure 5a,b (values are divided in two ranges of atomic number to better visualization). The same values are presented in Tables A4 and A5 (in Section VII). In Figure 5c,d, these values are normalized to the respective "DF" values. Figure 5 indicates that the partial fluorescence yields K-M 2 and K-M 3 from "DF" values are higher than the respective "DF,DS" values for all atomic numbers. These values are in higher disagreement at low atomic numbers, where at Z = 18 the relative difference is around 35%. At higher atomic numbers the agreement improves, and at Z = 95 the difference is lower than 3%.

Semi-empirical and empirical K-M 2 and K-M 3 partial fluorescence yield values
In Figure 6a,b, the semi-empirical and empirical partial fluorescence yields K-M 2 and K-M 3 are compared (the values are divided in two ranges of atomic number to better visualization). The same values are presented in Tables A9  and A10. These values are obtained from the combination of semi-empirical or empirical K-shell fluorescence yield values with Salem et al. transition probability values (as presented in Section 3.Q). From these figures and tables, it is clear that the semi-empirical and empirical partial Scofield's radiative Dirac-Fock calculations. [18] "EADL": values obtained from the Evaluated Atomic Data Library. [12] [Colour figure can be viewed at wileyonlinelibrary.com] FIGURE 6 "Bambynek1972": Values from Bambynek et al. earlier semi-empirical fit. [22] "Krause1979": values from Krause's semi-empirical fit. [40] "Bambynek1984": values from Bambynek et al. later semi-empirical fit. [41] "NIST": semi-empirical fit values presented in NIST's Fundamental Parameters Database. [28] "Kahoul": values from Kahoul et al. empirical fit. [29] "Daoudi": values from Daoudi et al. empirical fit. [30] fluorescence yields exhibit significantly high differences between each other, especially at low atomic numbers. In the range 3 ≤ Z < 20 some references present 10% to 100% difference to other references. As the atomic number increases better agreement is observed between all references. In the range 20 < Z < 80, the values are comprised within 10% difference from each other, and in the range 80 < Z < 96, differences are lower than 2%.
The K-M 2 and K-M 3 semi-empirical and empirical values are normalized with the respective "DF" theoretical values in Figure 6c,d, and with the respective theoretical "DF,DS" values in Figure 6e,f. These comparisons highlight that the empirical and semi-empirical values are higher than "DF,DS" values for almost all atomic numbers. Better agreement is found when comparing with the "DF" values. Table 7 presents the average relative difference between each semi-empirical or empirical K-M 2 values and the "DF,DS" and "DF" values. In Table 8 the same is presented, but regarding K-M 3 values. From these Tables, we infer that all semi-empirical and empirical values are in closer agreement with "DF" values. empirical fit. [30] [Colour figure can be viewed at wileyonlinelibrary.com]

CONCLUSIONS
The comparisons between all theoretical, empirical, and semi-empirical K-shell fluorescence yield values, K , from the different references exhibit that NIST's Fundamental Parameters Database values disagree with the other references, especially in the atomic number range 60 ≤ Z ≤ 80. As such, we suggest a change to values obtained from other references. The K , K-L 2 , K-L 3 , K-M 2 , K-M 3 partial fluorescence yields values from EADL and from Dirac-Fock/Dirac-Slater calculations ("DF,DS") exhibit some disagreement even though they were obtained from the same references. Pia et al. [13][14][15] had already pointed out this disagreement in their comparisons of partial fluorescence yields (normalized without accounting for radiationless transitions). Part of this disagreement can be due to the fact EADL includes a correction to the radiationless transition probabilities that we do not included in the present work.
Pia et al. work [13][14][15] highlighted that partial fluorescence yields (normalized without accounting for radiationless transitions) (NA) i obtained from Scofield's Hartree-Fock work [18] are in better agreement with the experimental values of Salem et al., [20] and as such, EADL should adopt Scofield's Hartree-Fock values. In the present work, the comparison consisted in partial fluorescence yields ij that where normalized accounting for radiationless transitions, as are presented in EADL. The comparisons from the present work highlight that regarding K , K-L 2 and K-L 3 values, when comparing the values obtained from using Scofield's Dirac-Slater calculations "DF,DS", and the values obtained using Scofield's Dirac-Fock calculations "DF", the "DF,DS" values are actually in better agreement with most of the semi-empirical and empirical values. The exception to the previous statement is the case of Bambynek et al. 1984 semi-empirical values, which are in excellent agreement with "DF" values. As for the K-M 2 and K-M 3 partial fluorescence yields, the "DF" values are in better agreement with all semi-empirical and empirical values. The comparisons in the present work do not support that the change in the EADL library, regarding what Scofield's values are adopted, will make EADL's partial fluorescence yield values in better agreement with the semi-empirical and empirical values available in literature. As such, we suggest that such a change in EADL should be backed up with further studies.
The present work introduced a method to obtain partial fluorescence yields K-L 2 , K-L 3 , K-M 2 and K-M 3 from the combination of semi-empirical or empirical fitting functions with Salem et al. [20] ratios (as described in Section 3.Q). This method can be a viable alternative of obtaining comprehensive values for data libraries, covering a large range of atomic number and a large range of transitions, as an alternative to relying in the existing theoretical data. However, some of the existing empirical and semi-empirical values present significant deviations between each other, especially at low atomic numbers, where differences around 100 % can be found between some of the references. In the range Z > 20 the deviations between the different references are comprised within 10% to a few percent.
From the comparisons of the present work, we concluded that the "DF" values are in excellent agreement with Bambynek et al. 1984 semi-empirical values. In fact, no pair of references presented better agreement.